{
  "id": 82791,
  "title": "training is unstable",
  "url": "/competitions/vsb-power-line-fault-detection/discussion/82791",
  "author_name": "",
  "post_date": "2019-03-04T11:14:56.361251300Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I was using RNN model, training/valid/test results are very unstable. This may due to the small training samples (only 2904 after converted). Is there some way to handle the instable results?</p>",
  "messages": [
    {
      "id": "483244",
      "postDate": "03/04/2019 11:14:56",
      "content": "<p>I was using RNN model, training/valid/test results are very unstable. This may due to the small training samples (only 2904 after converted). Is there some way to handle the instable results?</p>",
      "rawMarkdown": "I was using RNN model, training/valid/test results are very unstable. This may due to the small training samples (only 2904 after converted). Is there some way to handle the instable results?",
      "votes": null
    },
    {
      "id": "483638",
      "postDate": "03/04/2019 22:54:49",
      "content": "<p>Ensemble, better CV, augmentation, deal with outliers.</p>",
      "rawMarkdown": "Ensemble, better CV, augmentation, deal with outliers.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 483638,
      "author_name": "mpekalski",
      "author_url": "",
      "post_date": "03/04/2019 22:54:49",
      "content": "<p>Ensemble, better CV, augmentation, deal with outliers.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "483244": "I was using RNN model, training/valid/test results are very unstable. This may due to the small training samples (only 2904 after converted). Is there some way to handle the instable results?",
    "483638": "Ensemble, better CV, augmentation, deal with outliers."
  },
  "source": "meta"
}